Claude, GPT and DeepSeek in one place.
Switch models mid-conversation, set how hard they think, and see what each answer cost.
Invite only. Redeeming a code creates your account.
Switch models without starting over
Start somewhere cheap, move to Opus when the problem gets hard, then ask a second model to check the answer — all inside one conversation. Nothing is overwritten: regenerating a turn keeps the old one beside it.
Decide how hard it thinks
Reasoning effort is a dial, not a checkbox. Set it per message, from a quick pass to an exhaustive one. Models that don't offer a level say so rather than quietly rounding your choice to something else.
See what you spent
Every answer carries its own token count and credit cost. No monthly surprise, no guessing which conversation was the expensive one.
Have a code?
Redeeming it creates your account and grants its credits.
Open Inversify